deployments-and-agents
Build, adapt, validate, and troubleshoot ZenML pipeline deployments, model or agent services, DeploymentSettings, DockerSettings packaging, deployer selection, service discovery, and credential-free ML or LLM workflow smoke patterns.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- deployments-and-agents
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Build, adapt, validate, and troubleshoot ZenML pipeline deployments, model or agent services, DeploymentSettings, DockerSettings packaging, deployer selection, service discovery, and credential-free ML or LLM workflow smoke patterns.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/agent-and-deployment-workflows.md8.5 KB
- 参考文档references/deployment-settings.md8.2 KB
- 参考文档references/troubleshooting.md8.9 KB
- 脚本scripts/validate_deployment_settings.py7.1 KB
03 SKILL.md 正文
ZenML Deployments and Agents
Use this sub-skill when a task mentions ZenML pipeline deployments, deployed HTTP services, real-time inference, serving ML models, deployed agents, LLM or agent examples, DeploymentSettings, deployment DockerSettings, deployers, model deployers, services, dashboard/static UI files, deployment invocation, or credential-free adaptation of production examples.
Route First
- Read Agent and deployment workflows to choose between pipeline deployments, model-deployer/service workflows, agent examples, deterministic fallbacks, and production-readiness checks.
- Read Deployment settings before editing
DeploymentSettings,DockerSettings, CORS, secure headers, endpoint paths, dashboard files, app extensions, or deployer-specific settings. - Read Troubleshooting before running examples, changing stack/deployer configuration, debugging local-vs-remote behavior, or interpreting missing API keys, Docker, registries, services, and cloud credentials.
- Run validate_deployment_settings.py with
--exampleor--json <settings.json>for safe schema-level validation without deploying anything.
Boundary Rules
- Own example-backed production model, agent, and service workflows; pipeline deployments;
DeploymentSettings; deployer/model-deployer/service usage; HTTP invocation shape; embedded UI patterns; credential-free smoke adaptation; and deployment-specific troubleshooting. - Route low-level stack component implementation, integration flavor imports, service connectors, image builders, container registries, Docker daemon/backend issues, and cloud component registration to stacks-and-integrations.
- Route basic
@step,@pipeline, artifacts, materializers, schedules, hooks as authoring primitives, and local pipeline semantics to pipeline-authoring. - Route FastAPI server/store internals, REST routers, SQLModel schemas, migrations, RBAC, triggers, and server dependency wiring to server-and-stores.
- Route CLI/client resource audits,
zenml deployment list/describe/invokecommand behavior, login/connect, filters, secrets, and project/stack listing to cli-and-client. - Route repository test selection, formatting, linting, docs maintenance, and CI-equivalent checks to maintenance.
Working Pattern
- Classify the serving target first: pipeline deployment for a long-running HTTP service, batch/snapshot execution for one-off work, or legacy/specialized model deployer only when an integration explicitly requires model-server semantics.
- Make deployable pipelines accept explicit JSON-serializable parameters with defaults, return JSON-serializable outputs or stable artifacts, and keep heavyweight models/resources loaded through startup hooks or deployment-service state.
- Keep agent workflows credential-aware: use deterministic fallbacks, stubs, or skipped branches when API keys, network, LLM providers, Docker, services, or cloud credentials are unavailable.
- Treat
DeploymentSettingsas ASGI application configuration andDockerSettingsas image/runtime packaging; validate both before changing stack/deployer infrastructure. - Prefer local, credential-free checks first: static code review,
py_compile, settings validation, import checks, and small synthetic payloads. Deploy, invoke, start services, build images, or call LLMs only after the user explicitly authorizes the environment and credentials.
Safe Validation
python scripts/validate_deployment_settings.py --helpshows supported validation modes.python scripts/validate_deployment_settings.py --exampleprints and validates a tiny, credential-free deployment plus Docker settings payload.python scripts/validate_deployment_settings.py --json deployment_settings.jsonvalidates a user-supplied JSON payload containingdeploymentand/ordockersettings without loading source hooks, running pipelines, starting servers, using Docker, or reading credentials.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).